Revisions suggest placing low weight on quarterly changes in GDP

29 Jul 2011 by Jim Fickett.

Today the Bureau of Economic Analysis, in addition to publishing its first estimate for the latest GDP figures, through Q2, also released the routine revision of the previous two quarters, and the annual revision, for data going back to 2003. Since BEA helpfully took the trouble to graph some of the results, we can see at a glance just how much difference the revisions made. Here are the quarter-to-quarter changes in GDP, pre- and post-revision:

There are several very notable differences. The depth of the recession now looks considerably worse. Note especially that Q2 of 2010 was revised from just under 2% to just under 4% – about a doubling! And Q1 of this year was revised down from nearly 2% to barely positive.

To me the message is clear – despite the huge effort many people put into forecasting and interpreting the quarterly changes, they are, in fact, so noisy as to be of rather limited value. Certainly it matters if the quarterly change, annualized, is +2% rather than -7%. But when commentators get into detailed analysis of half a percent, the discussion can safely be ignored.

Here is one more graph that makes the same point. Gross Domestic Income should, absent measurement errors, be the same as Gross Domestic Product. And in fact, on an absolute scale, they come out pretty close. But small changes in the level make for big changes in the quarter-to-quarter percent change, and this graph shows, again, that one should not take the quarterly change in either one too seriously.